Scientometrics https://doi.org/10.1007/s11192-018-2812-9

Disciplinary structures in Nature, Science and PNAS: journal and country levels Jielan Ding1,2 • Per Ahlgren1,3 • Liying Yang1 • Ting Yue1,2 Received: 30 January 2018 Ó Akade´miai Kiado´, Budapest, Hungary 2018

Abstract This paper analyzes, using Web of Science publications and two time periods (2004–2006 and 2014–2016), the disciplinary structures in the three prestigious journals Nature, Science and PNAS, compared with two baselines: Non-NSP_Multi (multidisciplinary publications that have other source journals than Nature, Science and PNAS), and Non-Multi (publications assigned to other categories than Multidisciplinary). We analyze the profiles at two levels, journal and country. The results for the journal level show that for Nature and Science, the publications are considerably less concentrated to certain disciplines compared to PNAS. Biology is the dominant discipline for all the three journals. Nature and Science have similar publication shares in Medicine, Geosciences, Physics, Space science, and Chemistry. The publications of PNAS are highly concentrated to two disciplines: Biology and Medicine. Compared with Non-NSP_Multi and Non-Multi, the shares of Biology in NSP journals are higher, whereas the share of Medicine is lower. At the country level, 14 countries are included, among them the five BRICS countries. With respect to the NSP journals, the emphasis disciplines (in terms of world share of publications) of most countries other than USA are the disciplines in which USA has its weakest performance. The disciplinary structures of USA and of most of the other studied countries therefore tend to be different. Regarding Non-NSP_Multi and Non-Multi, the shapes of the disciplinary structures of the 14 countries can be roughly grouped into three groups, while there are more types of shapes for the countries in the NSP journals. For all five units of analysis, the discipline structures of most countries generally change only slightly between different time periods. The structures of some BRICS countries, however, change to a relatively large extent. Keywords Country  Disciplinary structure  Nature  Science  Proceedings of the National Academy of Sciences of the United States of America (PNAS)  Publication volume

& Liying Yang [email protected] 1

National Science Library, Chinese Academy of Sciences, Beijing 100190, China

2

University of Chinese Academy of Sciences, No. 19A Yuquan Road, Beijing 100049, China

3

School of Education and Communication in Engineering Sciences (ECE), KTH Royal Institute of Technology, Stockholm 10044, Sweden

123

Scientometrics

Introduction Activities related to the science system is an important issue in bibliometric research. As time goes on, the science system is not only growing in size but its structure is also changing (Van den Besselaar and Heimeriks 2001). The structure of the science system refers to the classification of the science system into different disciplines according the similarity of research (Wang 2016), and we use the term ‘‘disciplinary structure’’ to refer to this structure. The concept of disciplinary structure is used in quite much earlier research (e.g. Kozlowski et al. 1999; Yang et al. 2012; Li et al. 2015), a concept that has meaning similarities with the concepts disciplinary profiles (e.g. Bongioanni et al. 2014; MoyaAnego´n and Herrero-Solana 2013; Li 2017), disciplinary strengths (e.g. King 2004), disciplinary specializations (e.g. Radosevic and Yoruk 2014; Li 2017), research structure (e.g. Vinkler 2018) and science mapping (e.g. Boyack et al. 2005). Disciplinary structure is sometimes illustrated with disciplinary maps (e.g. Boyack et al. 2005; Leydesdorff and Rafols 2009; Van Eck and Waltman 2010) or radar charts (e.g. Gla¨nzel 2000; King 2004; Gla¨nzel et al. 2008) in bibliometrics. The study on disciplinary structure helps research and development (R&D) managers of a country or administrators of an institution to understand the profiles of their relevant environment, and provides them with necessary information for the science and technology (S&T) policy making process. There are many topics on disciplinary structure that are discussed in earlier literature, such as the accuracy of the structure (Boyack et al. 2005; Leydesdorff and Rafols 2009), techniques for mapping structure (Moya-Anego´n et al. 2004; Leydesdorff and Rafols 2009; Waltman et al. 2010; Waltman and Van Eck 2012), structural similarity (Bongioanni et al. 2014; Yang et al. 2012), interdisciplinary (Porter et al. 2008; Porter and Rafols 2009), balance of structure (Kozlowski et al. 1999; Yang et al. 2012; Li 2017), and disciplinary layout (e.g. Kozlowski et al. 1999; King 2004; Gla¨nzel et al. 2008; Yang et al. 2012; Aksnes et al. 2014; Radosevic and Yoruk 2014; Li 2017; Vinkler 2018). This paper deals with the disciplinary layout and the similarity of disciplinary structure at two levels, journal and country. For the latter level, there are many related, earlier studies reporting important findings. Kozlowski et al. (1999) studied the disciplinary structure of the science of post-communist countries of Central and Eastern Europe (CEE) and found that the inherited disciplinary structure of the science of post-communist countries of CEE carries strong common features of its past. For instance, the communist heritage is present in a relatively homogeneous research profile. King (2004) studied the national strengths in different disciplines for the G8 countries and revealed some asymmetries among the countries, such as a complementarity in citation impact between Germany and the UK, with German strengths in physical sciences and engineering complementing UK strengths in medical, life and environmental sciences. Gla¨nzel et al. (2008) studied 10 countries and found four basic paradigmatic patterns in publication profiles. Two of these patterns are the western model pattern, i.e., the characteristic pattern of the developed Western countries with clinical medicine and biomedical research as dominant fields, and the pattern of the former socialist countries, the present economies in transition and China with pronounced activity in chemistry and physics and less activity in the life sciences. Yang et al. (2012) compared the disciplinary structure of the G7 countries (representing high S&T level countries) and the BRICS countries—Brazil, India, Russia, and South Africa—and found that the life sciences play an important role for the G7 countries, while the BRICS countries focus on physics, chemistry, mathematics and engineering.

123

Scientometrics

Bongioanni et al. (2014) studied the dynamics of disciplinary structures of European countries and found that there is a convergence toward a unique European disciplinary profile of the scientific production, even if large differences in the scientific profiles still remain. It was also found that developing countries are converging toward the European model, while some developed countries are departing from it. Radosevic and Yoruk (2014) explored the changing role of world regions (North America, EU15, South EU, CEE, Former-USSR, Latin America, Asia Pacific and the Middle East) in disciplinary specializations and found that that the science systems are mostly characterized by strong inertia and historically inherited (dis)advantages. Asia Pacific, Latin America and CEE show strong catching-up characteristics but largely in the absorptive capacity of science. Aksnes et al. (2014) studied the research profiles of Netherlands and China using the Relative Specialization Index (RSI) and found that China has a scientific specialization, which deviates considerably from the world average, with a strong emphasis on research in engineering and the physical sciences. Li (2017) studied the disciplinary structure of 45 countries across 27 disciplines and found that while there has been a continuous process of convergence in national research profiles, nations differ greatly in their evolutionary patterns. Several scientometric indicators of 29 countries according to main scientific fields (Life, Natural, Applied, and Agricultural Sciences) were studied by Vinkler (2018), who distinguished four types of research structure by means of cluster analysis: countries preferring Life Sciences, with higher or lower share in Natural Sciences, and countries preferring Natural Sciences, with higher or lower share in Life Sciences. The findings of country level disciplinary layout reported in the preceding paragraphs have mainly been based on all journals covered by Web of Science (WoS). However, in this study we examine if the findings based on all WoS journals also apply to high prestige journals, and we restrict our attention to the three journals Nature, Science and Proceedings of the National Academy of Sciences of the United States (PNAS). The NSP journals are interdisciplinary scientific journals with very high prestige, publishing important and remarkable research across all fields, which are also described in their goals/aims. The aim of Nature is to publish the finest peer-reviewed research in all fields of science and technology on the basis of its originality, importance, interdisciplinary interest, timeliness, accessibility, elegance and surprising conclusions.1 Science seeks to publish those papers that are most influential in their fields or across fields and that will significantly advance scientific understanding.2 PNAS publishes brief first announcements of NAS members’ and foreign associates’ more important contributions to research and of work that appears to a member to be of particular importance.3 Many researchers across all fields focus highly on the three journals, the publications of which are often taken as representations of research with extraordinary high academic impact.4 Because of the importance and reputation of the three journals in scientific communication, fairly much research has investigated the properties of them, such as the internationality of them (Kaneiwa et al. 1988), their publication patterns and citation impact (Braun et al. 1989), the dynamic usage history of them (Wang et al. 2014), their collaboration characteristic (Ding and Rousseau 2015; Rousseau and Ding 2016; Xie et al. 2018) 1

http://www.nature.com/nature/about/.

2

http://www.sciencemag.org/about/mission-and-scope?_ga=2.115635599.881174117.1524540131-135903 6387.1524540131.

3

http://www.pnas.org/page/authors/purpose-scope.

4

For editorial process/policy, publication frequency, etc., we refer the reader to the web sites of the three journals.

123

Scientometrics

and their document types (Ding et al. 2016). However, and to our knowledge, less research has been devoted to the study of the properties of PNAS, compared to Nature and Science. With this study, we contribute to filling this gap. In view of the prestige Nature, Science and PNAS have within the scientific community, and the important role they play in various research assessments, we believe it is of importance to study them separately. Further, with respect to country level analyses, several of the studies mentioned above (e.g. Kozlowski et al. 1999; King 2004; Gla¨nzel et al. 2008; Yang et al. 2012; Aksnes et al. 2014; Radosevic and Yoruk 2014; Li 2017; Vinkler 2018) found that the disciplinary structures of high S&T level countries, such as USA, are different from the corresponding structures of fast-breaking countries, such as China. In this study, we do similar comparisons between country disciplinary structures of both types of countries. We selected the 10 most productive countries, regarding fractional publication output in Nature, Science and PNAS (taken as a totality). It turned out that the 10 most productive countries are high S&T level countries, with China as an exception. We included the other (relative to China) four BRICS countries, representing fast-breaking countries with relatively low S&T levels, as cases for comparison. We acknowledge, though, that this category is heterogeneous in some respects. For instance, China is the second most publishing country in the world and thereby deviates considerably from the other four countries in this respect. Further, we do not use the category in its primary sense—countries having a similar geopolitical position—but rather in a sense that focuses on rapidly occurring changes and S&T level. In this study, we treat the disciplinary structure of the three journals at a macro level of science, using, as have been indicated above, two levels of analysis: journal and country. We further take two sets of WoS publications, sets that act as baselines, into account, in order to obtain a comprehensive picture of the disciplinary structure of the three journals. We focus on the sciences, including social sciences, and we therefore do not include publications belonging to the humanities in the study. The remainder of this paper is structured as follows. In second section, the data, indicators and methods of the study are described. The third section gives the results, as well as interpretations of them. In the final section, the results are discussed and conclusions are put forward.

Data and methods The data source of the study was Bibmet, the bibliometric version of WoS at KTH Royal Institute of Technology (Sweden). Bibmet covers SCIE, SSCI and A&HCI, and contains about 51 million publications, with the earliest publication year equal to 1980, and is updated quarterly. To detect potential changes in disciplinary structure for journal level and country level, with regard to publications in Nature, Science and PNAS (and two baseline publication sets, described below), the two publication periods 2004–2006 and 2014–2016 were used. We set the period length to three years as we wanted the length to be long enough to have reasonably large publication volumes per period. The 10-year difference between the two time periods was selected after some experimentation. For instance, for small differences, like a three-year difference, only minor changes in disciplinary structure were observed. We concluded, based on the experiments that a 10-year difference seems to be a balanced one. We further think that one comparison period is enough regarding our focus

123

Scientometrics

period, 2014–2016, even if additional periods would have given rise to more information on disciplinary structure changes. In the end of September 2017, all publications (from the two periods) of the document types Article and Review and associated with at least one WoS subject category were retrieved from Bibmet, 7,590,245 publications in total. Since most publications of the types Article and Review have cited references, and we needed cited references (and citing publications) for re-classifying (see below) publications appearing in multidisciplinary journals, we restricted our dataset to publications of these two types. We removed all publications belonging to the humanities, where publications with source journals that have been assigned to at least one WoS subject category belonging to A&HCI were considered as belonging to the humanities. The rationale for removing publications belonging to the humanities is that the classification scheme used in the study (cf. next section) does not cover the humanities. In Bibmet, however, publications in multidisciplinary journals have been re-assigned to another WoS subject category than Multidisciplinary, based on the WoS subject categories of their cited references and citing publications. This yields, for instance, that if a Science publication has been re-assigned into an A&HCI subject category, the publication is removed. After the removal, 7,363,753 publications remained.

Classification of publications The study of disciplinary structure is not only based on manually constructed classification systems of science provided by certain databases, like WoS and Scopus, but also based on algorithmically constructed classification systems. Earlier research has investigated disciplinary structure on the basis of the former type of systems (e.g. Gla¨nzel 2000; Gla¨nzel et al. 2008; Porter et al. 2008; Leydesdorff and Rafols 2009; Leydesdorff et al. 2013), but also on the basis of latter type (e.g. Boyack et al. 2005; Klavans and Boyack 2006; MoyaAnegon et al. 2007; Klavans and Boyack 2009; Rafols et al. 2010; Van Eck and Waltman 2010; Bo¨rner et al. 2012; Waltman and Van Eck 2012). An advantage of algorithmically constructed classification systems is that they are more suitable for micro level studies than manually constructed systems. However, for a macro level study, as the one reported in this work, there are drawbacks with the algorithmic approach. For instance, it is difficult to obtain proper labels for the broad categories associated with a system level with low granularity. In this study, we deal with disciplinary structure on the basis of a manually constructed classification scheme, namely the Essential Science Indicators (ESI) scheme. The scheme, with only 22 categories, gives a macro view of the fields of science. The granularity of the WoS subject category scheme, with its about 250 categories, is considerably higher than granularity of the ESI scheme. Since we wanted to study the three journals at a macro level of science, we regarded the ESI scheme as proper basis for the analysis. Among the 22 ESI categories, there are several categories that refer to biology and medicine. We mapped the categories related to biology and medicine to the two categories Biology and Medicine, respectively. The reason for this merging is that we are interested in properties of the three journals at a classification granularity level corresponding to a considerably smaller number of disciplines than 22. In view of this interest, and in view of the relationships between the merged ESI disciplines, we believe that the performed merging is reasonable. We recognize, though, that the merging comes at a price, as information is lost. The merging resulted in 15 broad categories, which we collectively refer to by the expression ‘‘C15’’. In Table 1, the corresponding mapping is given.

123

Scientometrics Table 1 Mapping of ESI 22 categories to C15 categories, and of C15 categories to major fields ESI 22

C15_ full_name

C15_Abbreviated_name

Major field

Agricultural sciences Biology & biochemistry

Agricultural sciences

AGRI

Life sciences

Biology

BIOL

Medicine

MEDI

Microbiology Molecular biology & genetics Plant & animal science Clinical medicine Immunology Neuroscience & behavior Pharmacology & toxicology Psychiatry/psychology Chemistry

Chemistry

CHEM

Geosciences

Geosciences

GEOS

Mathematics

Mathematics

MATH

Physics

Physics

PHYS

Space science

Space science

SPAC

Computer science

Computer science

COMP

Engineering

Engineering

ENGI

Environment/ecology

Environment/ecology

ENVI

Materials science

Materials science

MATE

Economics & business

Economics & business

ECON

Social sciences, general

Social sciences, general

SOCI

Multidisciplinary

Multidisciplinary

MULT

Natural sciences

Applied sciences

Social sciences –

The 22 ESI categories belong to a small number of broader fields according to the research of Vinkler (2018) and Radosevic and Yoruk (2014). Vinkler (2018) grouped the ESI categories into five major fields: Social Sciences (Social Sciences, general, Economics & Business and Multidisciplinary), Natural Sciences (Chemistry, Physics, Geosciences, Mathematics, and Space Science), Applied Sciences (Engineering, Materials Science, Environment/Ecology, and Computer Science), Agricultural Sciences (Plant and Animal Science, Agricultural Sciences) and Life Sciences (the remaining categories). Radosevic and Yoruk (2014) did a similar grouping, in which the ESI categories, except Multidisciplinary, were assigned to four major fields: Social Sciences, Fundamental Sciences, Applied Sciences and Life Sciences (which also included all the fields in Vinkler’s Agriculture Sciences). In Table 1, we have mapped the C15 categories to four major fields, based on the grouping of ESI categories performed by Vinkler (2018). However, even if we treat disciplinary structure at a macro level in this study, we consider the resulting fields to be too broad for our purposes. For the assignment of publications to C15 categories, we used a mapping table (MT) in which journals indexed in SCI-EXPANDED and SSCI are assigned to C15 categories via ESI categories. The mapping from journals to ESI categories was obtained from InCites (provided by Clarivate Analytics).5 The MT data used were ISSN, EISSN and full journal title. The assignment was executed in three-steps, described in the following list: 5

Bibmet does not include the ESI scheme.

123

Scientometrics

1.

2.

3.

Matching of the ISSN of the source journal of the publication against the ISSN column of MT. If a match occurred, the publication was assigned to the corresponding C15 category. If no match was obtained, step 2 below was executed. Matching of the ISSN of the source journal of the publication against the EISSN column of MT. If a match occurred, the publication was assigned to the corresponding C15 category. If no match was obtained, step 3 below was executed. Matching of the full title of the source journal of the publication against the full title column in MT. For this matching, an approach based on Levenshtein distance (LD) (Levenshtein 1966) was used. The LD between two strings of symbols can informally be defined as the minimal number operations that have to be performed in one of the strings in order to make them identical, where the admitted operations are deletion of a symbol, insertion of a symbol and substitution of a symbol by a symbol. Let Lmax(s1, s2) be the maximal length with respect to the strings s1 and s2. We define the normalized Levenshtein similarity between s1 and s2, LSnorm(s1, s2), as 1 - (LD(s1, s2)/Lmax(s1, s2)). Now, if LSnorm between the full title of the journal of a publication and a corresponding MT title was greater than 0.95, we regarded this as a match, and the publication was assigned to the corresponding C15 categories.

After the three-step assignment procedure, which we denote by ‘‘AP’’, 7,071,077 publications remained (were assigned to C15 categories), corresponding to 96% of the 7,363,753 non-humanistic publications.

Re-classification of publications appearing in multidisciplinary journals All publications among the 7,071,077 that were assigned to the C15 category. Multidisciplinary in the AP, 183,897 publications, were re-classified on the basis of the C15 categories of their cited references and of the publications that cite them, and by use of the mapping table MT. We used, then, not only cited references but also citing publications for the re-classification task. Our approach obviously uses more information compared to approaches in which only cited references are used for the re-classification (e.g. Gla¨nzel and Schubert 2003; Gla¨nzel et al. 1999a, b). Let MDISC0 stand for the set of the 183,897 publications. For such a re-classification approach to work, a publication in MDISC0 must have (a) at least one cited reference that points to a publication covered by SCI-EXPANDED or SSCI, or (b) at least one citing publication that is covered by SCI-EXPANDED or SSCI. If a publication in MDISC0 does not satisfy (a) or (b), none of its cited references/citing publications (if any) are recorded in SCI-EXPANDED or SSCI. In such a case, clearly, the publication could not be assigned to a C15 category based on its citation links, and the publication had to be deleted from the study. It turned out that 181,458 of the 183,897 MDISC0 publications satisfied condition (a) or condition (b) (or both). Let MDISC1 be the set of these publications. The assignment procedure AP was then applied a second time to the MDISC1 publications. However, in this second iteration, the cited references and citing publications of the MDISC1 publications were matched to the data of the mapping table MT. For each MDISC1 publication, a list of its cited references/citing publications, together with the corresponding C15 categories, was generated. Note that such a list can be empty with respect to C15 categories (none of the items, i.e., cited references/citing publications, could be matched in AP to a C15 category via the source journal of the item) and that the list might have items assigned in AP to the C15 category Multidisciplinary. After

123

Scientometrics

removal of cited references/citing publications that could not be assigned to any C15 category or were assigned to the C15 category Multidisciplinary, a subset of MDISC1, say MDISC2, was obtained with 181,127 publications. For each publication in MDISC2, the most frequently occurring C15 category (which cannot, at this point, be identical to Multidisciplinary), regarding its list of cited references/citing publications, was assigned to the publication. In case of ties, i.e., in cases where two or more C15 categories (not including Multidisciplinary) categories represented in the list of a publication had the maximum frequency, the publication was deleted. In this way, a subset of MDISC2, say MDISC3, was obtained with 177,028 publications, which were re-classified to C15 categories (not including Multidisciplinary), about 96% of the 183,897 in MDISC0 (the set of publications that were assigned to the C15 category Multidisciplinary in the first iteration of AP). The four datasets involved in the re-classification procedure are briefly described in Table 2.

Journal level We compare the publications in Nature, Science and PNAS with two additional units of analysis, which act as baselines: multidisciplinary publications that have other source journals than Nature, Science and PNAS (Non-NSP_Multi), and publications assigned in AP (first iteration) to other C15 categories than Multidisciplinary (Non-Multi). For NonMulti, we removed the publications that did not satisfy the conditions (a) and (b) mentioned in the second paragraph of the preceding section. This was done in order to avoid bias. It might be the case that publications in MDISC0 not satisfying (a) or (b), and therefore deleted from the study, tend to be of a certain kind. For instance, such publications might to a large extent belong to the social sciences. If this would be the case, it would obviously not be a good idea to keep them in one of the baselines of the study. After the removal, 6,790,769 Non-Multi publications remained. We finally arrived at 6,790,769 (the number of Non-Multi publications) ? 177,028 (the number of publications in MDISC3) = 6,967,797 publications (about 95% of the 7,363,753 non-humanistic publications). All these were used at the journal level of analysis. Table 3 reports the number of publications by unit of analysis, i.e., Nature, Science, PNAS, Non-NSP_Multi and Non-Multi, time period, and by number of remaining publications after certain methodological steps described above. For the latter, the columns ‘‘C15_cat’’ gives the number publications with a C15 category after the first iteration of the Table 2 Datasets involved in the re-classification procedure Dataset

No. of publ.

Description

MDISC0

183,897

Publications assigned to Multidisciplinary in the first iteration of AP based on their source journal

MDISC1

181,458

Publications with no cited references/cited publications covered by SCIEXPANDED or SSCI removed from MDISC0

MDISC2

181,127

Publications such that each of their cited references/cited publications could not be assigned to a C15 category or was assigned to the category Multidisciplinary removed from MDISC1

MDISC3

177,028

Publications with two or more maximum frequencies with respect to the C15 categories of their cited references/citing publications removed from MDISC2

123

2,589,567

Total

2,542,707

2519,552

23,155

7958

9575

2719

2903

2,542,000

2,519,552

22,448

7370

9498

2705

2875

2,524,000

2,501,980

22,020

6999

9497

2701

2823

The numbers mentioned in the text are given in italics in the table

2565,016

9316

Non-NSP_Multi

Non-Multi

9575

PNAS

24,551

2727

Total-Multi

2933

Science

adress

4,481,510

4,322,164

159,346

144,260

9987

2462

2637

C15_cat

re_class

C15_cat

cited/citing

2014–2016

2004–2006

Nature

UoA

4,429,520

4,271,217

158,303

143,225

9985

2460

2633

cited/citing

4,425,797

4,271,217

154,580

139,737

9818

2425

2600

re_class

4,417,256

4,262,866

154,390

139,563

9815

2414

2598

adress

7,071,077

6,887,180

183,897

153,576

19,562

5189

5570

C15_cat

Total

6,972,227

6,790,769

181,458

151,183

19,560

5179

5536

cited/citing

6,967,797

6,790,769

177,028

147,107

19,316

5130

5475

re_class

6,941,256

6,764,846

176,410

146,562

19,312

5115

5421

adress

Table 3 Number of publications by unit of analysis, time period, and number of remaining publications after certain methodological steps. ‘‘UoA’’ stands for ‘‘unit of analysis’’

Scientometrics

123

Scientometrics

procedure AP, while the columns ‘‘cited/citing’’ gives the number of publications with at least one cited reference, or at least one citing publication (or both). The columns ‘‘re_class’’ reports the number of publications with a unique highest frequency C15 category (C15 categories other than Multidisciplinary) after the second iteration of AP (note that this does not pertain to Non-Multi). The ‘‘address’’ columns concern the country level (cf. next section). In this paper, we operationalize disciplines as C15 categories. In order to map the disciplinary structure of Nature, Science and PNAS across time, publication counts is used to obtain the relative contribution of the publications in a given discipline in a given time period, with respect to all publications (obtained in the study) in the journal in the time period. The same approach is used for the two baselines Non-NSP_Multi and Non-Multi. Thus, for unit of analysis i, discipline j, and time period t, we define Pijt as the number of publications in unit of analysis i in discipline j in time period t, and Rijt as the percentage of publications of discipline j in unit of analysis i in time period t relative to the total number of publications in unit of analysis i in time period t. Formally: Rijt ¼

Pijt  100 Pit

ð1Þ

where Pit is the total number of publications in unit of analysis i in time period t. Then, for unit of analysis i in time period t, the (relative) disciplinary structure is represented by the vector Dit, defined as Dit ¼ ðRi1t ; Ri2t ; . . .; Rimt Þ

ð2Þ

where m is the number of disciplines for unit of analysis i in time period t. These vectors were used to detect potential changes in disciplinary structure across the two periods and by unit of analysis. At the journal level, the sum of the components of a vector Dit clearly equals 100.

Country level For the country level, we started with the publication sets finally obtained at the journal level (Table 3, the columns ‘‘re_class’’). However, there exist publications without addresses among the 6,967,797 publications finally obtained at the journal level. After removal of such publications, 6,941,256 publications remained (Table 3, the columns ‘‘address’’), about 94% of the 7,363,753 non-humanistic publications. Due to co-authorship, a publication might involve more than one country among its addresses. Therefore, we make use of address fractionalization. If, for instance, a publication has two USA addresses and one France address, USA contributes with 2/3 and France with 1/3 to the publication. The reason that we use address fractionalization, and not the more proper author fractionalization, is that WoS does not include couplings between authors and their addresses before the publication year 2008. As indicated in Sect. 1, we include the 10 most productive countries, regarding fractional publication output in Nature, Science and PNAS (taken as a totality), the remaining 6,941,256 publications and in the publication period 2014–2016 in the analysis. In addition to the top 10 countries, we include the other (relative to China) four BRICS countries as cases for comparison, countries that are usually seen as developing countries in S&T level. These 14 countries are the ones we consider in this study, and which are represented in Table 4.

123

58.2

46.4

Australia

Netherlands

5.6

84.5

Switzerland

South Africa

63.5

Canada

10.2

110.4

France

Russia

72.1

Japan

7.7

110.5

China

3.6

196.9

Germany

India

241.6

Brazil

1279.3

UK

3.0

9.1

3.4

7.9

42.6

36.7

60.9

61.0

93.8

94.7

86.0

172.2

194.2

1273.4

13.5

23.4

40.0

35.6

151.9

166.1

160.3

258.6

338.4

393.9

445.0

523.4

628.2

5450.8

22.1

42.7

46.9

51.1

240.9

261.1

305.6

383.1

542.6

560.7

641.5

892.6

1064.0

8003.5

NSP

788.5

593.9

3873.0

3142.1

2950.0

4084.8

1817.0

4041.9

4209.6

7927.5

27,051.0

7352.2

7146.7

28,843.9

Non-NSP_Multi

21,746.0

72,988.8

151,723.7

97,701.3

66,704.5

114,770.5

43,328.7

126,106.7

132,877.7

180,327.5

701,310.4

196,529.2

195,187.8

869,923.4

Non-Multi

27

22

32

23

10

9

6

8

5

7

4

3

2

1

Nature

PNAS

Nature

Science

Rank

Number of publication fractions

USA

Countries

32

21

31

22

9

10

8

7

5

4

6

3

2

1

Science

32

25

22

23

10

8

9

7

6

5

4

3

2

1

PNAS

31

25

24

22

10

9

8

7

6

5

4

3

2

1

NSP

26

31

10

13

15

8

17

9

7

3

2

4

5

1

Non-NSP_Multi

31

15

6

13

17

11

21

10

9

5

2

3

4

1

Non-Multi

Table 4 Number of publication fractions and ranks of the top 10 most productive countries and the BRICS countries in the time period 2014–2016 by unit of analysis/NSP

Scientometrics

123

Scientometrics

For a country c, we define Pijtc as the number of c publication fractions in unit of analysis i in discipline j in time period t. We further define Rijtc as the percentage of c publication fractions in unit of analysis i in discipline j in time period t relative to P’ijt, the number of publications, with at least one address, in unit of analysis i in discipline j in time period t. Formally: Rijtc ¼

Pijtc  100 P0ijt

ð3Þ

Since fractionalization is used, the sum of the percentages across all countries for unit of analysis i, with regard to discipline j and in time period t, is equal to 100. For country c and in unit of analysis i in time period t, the (relative) disciplinary structure is represented by the vector Ditc, defined as Ditc ¼ ðRi1tc ; Ri2tc ; . . .; Rimtc Þ

ð4Þ

where m is the number of disciplines for unit of analysis i in time period t. In contrast to the journal level, the sum of the components of a vector Ditc is not necessarily equal to 100. Here it is important to have before the mind that a country’s publication fractions in a discipline are related to the total number of publications in the discipline, not to the total number of publications fractions of the country across the disciplines. So if a country c performs well in a discipline j, this means that c has a high share of the publications in j, compared to corresponding shares of c in other, considered disciplines. In this case, c’s international standing regarding publication volume is relatively good in discipline j. We used these vectors to compare the disciplinary structures of the countries in the period 2014–2016 by unit of analysis. Pairs of vectors–where the two vectors concern the same unit of analysis, the same time period, but two countries–were given as input to the cosine measure (Salton and McGill 1983). We thereby obtained disciplinary structure similarities, for each of the five units of analysis, between the two countries within a time period. Moreover, to detect potential changes in disciplinary structure across the two periods, again pairs of vectors—where the two vectors concern the same unit of analysis, the same country, but different time periods—were given as input to the cosine measure. We in this case obtained disciplinary structure similarities, for each of the five units of analysis, between the two time periods within a country. The similarity between two vectors, Ditc and Dit0 c0 , is defined as follows: Pm j¼1 Rijtc  Rijt0 c0 SimðDitc ; Dit0 c0 Þ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5Þ 2ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 Pm  Pm  0 c0 R R  ijtc ijt j¼1 j¼1 where t0 and c0 stand for a time period and a country, respectively.

Results Disciplinary structures at the journal level Figures 1, 2 and 3 show the disciplinary structures of Nature, Science and PNAS, respectively, for the periods 2004–2006 and 2014-2016 (the percentages above the bars apply to the period 2014–2016; the underlying data is shown in Appendix Table 6). Notice that the bars across the disciplines, for a given journal and a given time period, correspond

123

Scientometrics

Fig. 1 The percentage of publications across disciplines in Nature in the time periods 2004–2006 and 2014–2016

Fig. 2 The percentage of publications across disciplines in Science in time periods 2004–2006 and 2014–2016

Fig. 3 The percentage of publications across disciplines in PNAS in time periods 2004–2006 and 2014–2016

123

Scientometrics

to a vector of the type defined in Eq. (2). It is clear from the three figures that Biology is the dominant discipline for all the three journals, regardless of time period. Medicine is the second dominant discipline of the three journals, but the shares for Medicine are considerably lower compared the shares for Biology. Nature and Science have relatively high publication shares also in Geosciences, Physics, Space science, and Chemistry. The publications in PNAS are, though, highly concentrated to two disciplines: Biology and Medicine. In 2014–2016, about 79% of the publications belong to these two disciplines in PNAS. In Figs. 4, 5, the disciplinary structures of the publications in multidisciplinary journals other than Nature, Science and PNAS (the baseline Non-NSP_Multi), and of the publications in all (non-humanistic) and non-multidisciplinary WoS journals (the baseline NonMulti), for the two time periods are displayed (the percentages above the bars apply to the period 2014–2016; the underlying data is shown in Appendix Table 7). Compared with baselines Non-NSP_Multi and Non-Multi, the shares of Biology in NSP journals are much higher, especially compared to the Non-Multi baseline, which indicates the high weight of the Biology discipline in the three prestigious journals. The shares of Medicine in NSP

Fig. 4 The percentage of publications across disciplines in Non-NSP_Multi baseline in the time periods 2004–2006 and 2014–2016

Fig. 5 The percentage of publications across disciplines in Non-Multi baseline in the time periods 2004–2006 and 2014–2016

123

Scientometrics

journals is somewhat lower than the corresponding share in Non-Multi baseline, and much lower than the share in Non-NSP_Multi baseline. The emphasis on Medicine in the NSP journals is thus not as strong as it is in non-multidisciplinary journals and non-NSP multidisciplinary journals. Comparing the two baselines Non-NSP_Multi and Non-Multi, publications in Non-NSP_Multi baseline are much more associated with Medicine and Biology than publications in Non-Multi baselines. We also observe that the share of Social Sciences, general is substantially higher in Non-Multi baseline than in the other four units of analysis.

Disciplinary structures at the country level In Table 4 of Sect. 2, the 10 most productive countries–regarding fractional publication output in NSP and in the period 2014–2016–and the BRICS countries are represented, where the underlying dataset is the set of the remaining 6,941,256 publications (see the ‘‘Data and methods’’, the section ‘‘Country level’’). The countries are sorted descending after 2014–2016 NSP total publication output. Table 4 reports number of publication fractions and ranks by country and by unit of analysis/NSP for the period 2014–2016. The number of publication fractions of the 14 countries by unit of analysis/NSP for the period 2004–2006 is reported in Appendix Table 8. The publication volumes of some disciplines are small in NSP journals (Figs. 1, 2 and 3 and Appendix Table 9). It is not statistically meaningful to calculate country world shares for such disciplines. Therefore, we only took into account, for a given journal, the disciplines with more than 1% of the publications in the time period 2014–2016. In the remainder of this paper, we refer to such disciplines as main disciplines. This yielded 8, 9 and 7 main disciplines in Nature, Science and PNAS, respectively, at the country level. For the baselines Non-NSP_Multi and Non-Multi, we took into account each discipline that belongs to at least one of the main disciplines of Nature, Science and PNAS. This yielded nine disciplines for the two baselines. The main disciplines were used both in the betweencountry analysis (Sect. 3.2.1) and in the within-country analysis (Sect. 3.2.2). For the latter, we observed that main disciplines in NSP journals were quite stable across the two considered time periods.

Country disciplinary structure in the period 2014–2016 Figures 6, 7, 8, 9 and 10 visualize for the five units of analysis, in the form of radar plots, the 2014–2016 disciplinary structure of the countries. In the radar plots, the studied countries are sorted (row-wise) descending after NSP total publication output (cf. Table 4). Notice that a radar plot corresponds to a vector of the type defined in Eq. (4). The cosine similarity values between countries with regard to the five units of analysis and for the two considered periods are reported in Appendix Tables 9, 10, 11, 12, 13. Sample observations for the five units of analysis Nature (a) USA has a fairly balanced structure, which is not far from a circle structure with relatively more emphasis on Medicine, Materials Science, Chemistry, and Biology. The disciplinary structures of USA and Germany are similar, with a cosine similarity value of 0.953. China and Japan have similar disciplinary structures with emphasis on Chemistry and Physics (similarity value equal to 0.919).

123

Scientometrics

Fig. 6 The world share in main disciplines of the 14 countries in Nature in period 2014–2016. Note. The countries are sorted (row-wise) according to the country sequence in Table 4

123

Scientometrics

Fig. 7 The world share in main disciplines of the 14 countries in Science in period 2014–2016. Note. The countries are sorted (row-wise) according to the country sequence in Table 4

123

Scientometrics

Fig. 8 The world share in main disciplines of the 14 countries in period PNAS in 2014–2016. Note. The countries are sorted (row-wise) according to the country sequence in Table 4

123

Scientometrics

Fig. 9 The world share in main disciplines of the 14 countries in Non-NSP_multi baseline in period 2014–2016. Note. The countries are sorted (row-wise) according to the country sequence in Table 4

123

Scientometrics

Fig. 10 The world share in main disciplines of the 14 countries in Non-Multi baseline in period 2014–2016. Note. The countries are sorted (row-wise) according to the country sequence in Table 4

123

Scientometrics

(b) UK and Canada have similar disciplinary structures with emphasis on Environment/ Ecology and Geosciences (similarity value equal to 0.929). (c) Many examined countries focus strongly on Physics. Germany and Russia are such examples. The publication shares of Physics are the highest across all considered disciplines for these two countries. Also Japan and Switzerland have comparably large shares in Physics. (d) The plots for France and India show an irregular shape structure. Space Science has the highest share for these two countries. (e) Other countries with irregular shape structure plots are Australia, Netherland, Brazil and South Africa. For these countries, Environment/Ecology has the highest share of publications. Science (a) USA has a quite even disciplinary structure with relatively more emphasis on Social Sciences, general. USA and UK have similar disciplinary structures with a similarity value of 0.960. (b) According to cosine similarity values, China’s nearest neighbor is Switzerland. The similarity value is 0.910, and both countries have the highest publication shares in Materials Science. (c) The plots for Australia and Brazil have a similar, irregular shape structure (similarity value equal to 0.955), and both countries, like Canada, focus most on Environment/ Ecology. (d) Several countries have their highest publication shares in Medicine, Space Science, Physics or Chemistry. Netherlands and India have their highest shares in Medicine, Germany, Japan and France in Space Science, Germany, Japan and Russia in Physics, and China, Japan and Netherlands in Chemistry. (e) The plots for Russia and South Africa have irregular shape structures with different emphasis: The highest share for Russia concerns Geosciences, whereas the highest share for South Africa concerns Social Sciences, general. PNAS (a) USA has a fairly balanced disciplinary structure, and stands for the largest shares (47–65%) of the total publication fraction output across the countries for all included disciplines. UK and Netherlands also have a quite balanced structures. Their similarity values with USA is 0.98 and 0.97, respectively. (b) Several countries have their highest shares in Physics, e.g. China, Japan, Netherlands, India and Russia. (c) Several examined countries have their highest shares in Environment/Ecology, e.g. Canada, Australia, Brazil and South Africa. (d) The highest shares for Germany and France concern Geosciences, whereas Switzerland has its highest share in Chemistry. Non-NSP_multi (a) USA has a fairly balanced disciplinary structure, with relatively more emphasis on Space science, Biology, Social Sciences, general, Environment/Ecology and Medicine. UK, Canada and Australia have similar disciplinary structures compared to USA, with overlap in emphasis disciplines. The latter disciplines of those countries mainly belong to Life Sciences. (b) China has a different emphasis from USA with more focus on Natural Sciences and Applied Sciences such as Materials Science, Chemistry, Physics and Geosciences. (c) The emphasis disciplines for most countries, for example Germany, Japan, France and Switzerland, belong to both Life Sciences and Natural Sciences.

123

Scientometrics

(d) Netherlands, Brazil, India, Russia and South Africa have irregular shape structure plots. For these countries, two or three disciplines have much higher world shares compared to the other disciplines. Space science, Geosciences, Social Sciences, general, Environment/Ecology, Medicine, and Physics are the disciplines with the highest shares. Non-Multi (a) USA, UK, Canada, Australia and Netherlands are countries with relatively high world shares in Social Sciences, general, Medicine, Space Science and Biology. (b) China, Brazil, India have plots with opposite shape structures compared to the shape structures for the countries mentioned in point a) above. These three BRICS countries focus on Materials Science, Chemistry and Physics, and they therefore focus on Natural Sciences and Applied Sciences. (c) Germany, Japan, France, Switzerland have fairly balanced structures. The similarity values between these countries are close to 1. The emphasis is related to Life sciences disciplines and Natural Sciences/Applied Sciences disciplines. (d) The plots for Russia and South Africa have irregular shape structures, where two or three disciplines dominate. Russia focuses on Physics, Space Science and Geosciences, whereas South Africa focuses on Social Sciences, general and Environment/Ecology. Disciplinary structure in NSP journals versus disciplinary structure in the two baselines A comparison between disciplinary structures in NSP journals and such structures in the two baselines, Non-NSP_multi and Non-Multi, reveals that there is more variation in shape structure among the countries in NSP journals compared to the two baselines. More countries exhibit (different) irregular shape structures, with one or two disciplines that have much higher world shares than the other disciplines, in NSP journals than in the baselines. For the latter units of analysis, the shape structures can be roughly grouped into a few types. For the two baselines, especially for Non-Multi, the 14 countries can be divided into three groups according to discipline emphasis: (1) Countries with emphasis on disciplines in Life Science/Social Sciences. These countries are usually developed countries, like USA. (2) Countries with emphasis on disciplines in Natural Sciences/Applied Sciences, which are related to national infrastructure. In the baseline Non-NSP_multi, only China belongs to this group. In the baseline Non-Multi, three BRICS countries belong to the group. These countries are examples of fast-breaking countries with lower S&T level. (3) Countries with approximately equal emphasis on Life Science/Social Sciences and Natural Sciences/Applied Sciences. For instance, Germany, Japan, France, Switzerland belong to this group in both baselines. For NSP Journals, the 14 countries cannot easily be divided into groups based on discipline emphasis. However, a division between USA and the other studied countries can be made. The emphasis disciplines (in terms of world share of publications) of most countries other than USA are the disciplines in which USA has its weakest performance. In Nature, and for USA, Environment/Ecology, Physics and Geosciences are the three disciplines with the lowest world shares, whereas a majority of the other countries have a strong emphasis on Environment/Ecology and Physics. Similar patterns can be observed also for Science and PNAS. It can be concluded, then, that the disciplinary structures of USA on the one hand, and of most of the other studied countries on the other hand, tend to be different. From further inspection of the radar plots in Figs. 6, 7, 8, 9 and 10, we compare the world share of publications in NSP journals with the corresponding share in the two baselines, in the main disciplines, for the 14 countries and in the period 2014–2016 (the underlying data is shown in Appendix Table 14). It is evident that Germany, Netherlands,

123

Scientometrics

Switzerland, UK and USA—especially the latter three—are countries that perform better in most main disciplines, with regard to relative publication output, in highly visible research (taken as research reported in publications in NSP journals) than in overall research (taken as research reported in the publications of the two baselines). The other countries, however, perform worse in most main disciplines in highly visible research than in overall research. The five countries mentioned above rank in the top 1–8 in ‘‘The Global Competitiveness Index 2016–2017 Rankings’’ and rank in the top 1–12 in the ‘‘12th pillar: innovation capacity’’ in The Global Competitiveness Report 2016–2017 (World Economic Forum 2017). Those countries have a beneficial standing with respect to the indicator ‘‘Gross domestic expenditure on R&D (GERD) per capita at current prices and PPP’’ (OECD 2018), and they clearly have high S&T levels. These high levels are more advantageous in highly visible research, in which it is more difficult for a country to assert itself compared to overall research.

Disciplinary structure between two time periods within countries Table 5 gives the cosine similarity values for pairs of time periods within the 14 countries for the NSP journals, as well as for the two baselines. In the table, values below 0.8 are given in italics. For all five units of analysis, the discipline structures of most countries generally change only slightly between different time periods. The structures of some BRICS countries, however, change to a relatively large extent. For instance, the structures of Russia and South Africa regarding Science and PNAS. Taking Russia and Science as an example, the large differences between the two periods occur in Environment/Ecology (world share from 1.45 to 0.01%), Materials Science (world share from 0.00 to 0.19%), Social Sciences, general (world share from 2.5 to 0.0%) and Space Science (world share

Table 5 Similarity values between the time periods 2004–2006 and 2014–2016 regarding disciplinary structures within countries by unit of analysis Countries

Nature

Science

PNAS

Non-NSP_Multi

Non-Multi

USA

0.991

0.998

0.997

0.929

0.998

UK

0.977

0.946

0.951

0.818

0.995

Germany

0.915

0.972

0.906

0.905

0.992

China

0.621

0.913

0.718

0.770

0.975

Japan

0.774

0.965

0.909

0.843

0.977

France

0.982

0.908

0.992

0.911

0.994

Canada

0.736

0.908

0.923

0.706

0.995

Switzerland

0.917

0.911

0.962

0.814

0.991

Australia

0.989

0.773

0.950

0.849

0.984

Netherlands

0.779

0.938

0.913

0.919

0.987

Brazil

0.899

0.688

0.564

0.872

0.964

India

0.606

0.932

0.692

0.839

0.993

Russia

0.962

0.414

0.654

0.668

0.984

South Africa

0.823

0.774

0.418

0.820

0.968

The countries are sorted (row-wise) according to the country sequence in Table 4

123

Scientometrics

from 0.48 to 0.81%). However, when these differences are interpreted, the small publication volumes of Russia should be taken into account. The results show that the disciplinary structure is more stable in the Non-multi baseline than in the other units. For this unit of analysis, all similarity values are greater than or equal to 0.964.

Discussion and conclusions In this work, we have dealt with disciplinary structure in Nature, Science and PNAS, as well as in two baselines (Non-NSP_Multi: publications in non-NSP multidisciplinary journals; Non-Multi: publications in non-multidisciplinary journals). The investigation has been performed at two levels: journal and country. For the journal level disciplinary analysis, we have observed that Nature and Science are considerably less concentrated to certain disciplines compared to PNAS. Biology is the dominant discipline for all the three journals. Nature and Science have similar publication shares in Medicine, Geosciences, Physics, Space science, and Chemistry. In contrast to Nature and Science, the publications in PNAS are highly concentrated to two disciplines: Biology and Medicine. Compared with Non-NSP_Multi and Non-Multi, the shares of Biology in NSP journals are much higher, especially compared to Non-Multi, whereas the share of Medicine are lower, especially compared to Non-NSP_Multi. For the country level disciplinary analysis and with respect to the NSP journals, the emphasis disciplines (in terms of publication world share) of most countries other than USA are the disciplines in which USA has its weakest performance. The disciplinary structures of USA and of most of the other studied countries therefore tend to be different. Regarding the two baselines, the disciplinary structures of the 14 countries are not as diverse as for the NSP journals. Developed countries, such as USA, tend to have emphasis on Life Sciences and Social Sciences, while fast-breaking countries with lower S&T levels, such as China, tend to have emphasis on Natural Sciences and Applied Sciences. Similar findings have been reported in earlier research (e.g. King 2004; Gla¨nzel et al. 2008; Zhou and Glanzel 2010; Yang et al. 2012; Aksnes et al. 2014; Radosevic and Yoruk 2014; Jurajda et al. 2017; Li 2017; Vinkler, 2018). For all five units of analysis, the discipline structures of most countries generally change only slightly between different time periods. The structures of some BRICS countries, however, change to a relatively large extent. Earlier research has raised the following, important question: What are the major determinants of a country’s disciplinary structure? First, historical factors and path dependence are importance factors, in view of earlier research. Kozlowski et al. (1999) suggest that historical factors matter a lot in forming the disciplinary structures of countries. The strong homogeneity of the disciplinary structures of the post-communist countries is to a large extent due to their common communist legacy, according to Radosevic and Yoruk (2014), who also pointed out that science systems are characterized by strong inertia and path dependency in areas of historically inherited (dis)advantages. Second, strategy and policy in science and technology also matters to some extent. Radosevic and Yoruk (2014) state that Asia Pacific seems to follow a science policy, which prioritizes applied sciences (see, moreover and for instance, Wong 2013; Harzing and Giroud 2014), whereas South EU has opted for fundamental sciences. Li (2017) suggests that national differences in disciplinary profiles may be caused by direction of persistent investment (i.e., investment in certain fields of R&D) and by factor endowments (i.e., big country versus small country). Third, the development statues of a country also seem to affect disciplinary structure. Li (2017) argues that the level of development (developed country

123

Scientometrics

versus developing country), geographical location (continental location, whether adjacent to a large economy, etc.), and cumulative mastery of knowledge through national culture and traditions are related to disciplinary structure. The above-mentioned factors are related to a country itself, and we may describe them as self -related factors. However, it seems reasonable to assume that the disciplinary structure of a country is also influenced by its position in the international competition regarding disciplines. If other countries have taken the lead in the international competition in a discipline, it is difficult for the country to successfully compete in the discipline. Our results indicate that the competitiveness force factor is more important than the self-related factors for country disciplinary structure regarding the three prestigious NSP journals, since the emphasis disciplines of most of the other studied countries tend to be different from to the emphasis disciplines of USA. The self-related factors, though, might be more important factors for country disciplinary structure regarding the two baselines, because of the absence of the USA-other countries opposition tendency. Our results, then, indicate that self-related factors, such as historical factors, are more important in the overall research (taken as research reported in the publications of the two baselines) disciplinary structure, while in highly visible research (taken as research reported in publications in NSP journals), the competition between countries is more intense, and the advantages of USA in the competition are more obvious. Thus, for disciplinary structure in different level output, the relative contribution of the involved factors might be different. We think that this is a promising direction for future research (Li 2017). As suggested by one of the reviewers, the nationalities of journal board members might be a factor that influences the disciplinary structures of countries. This is reasonable idea, for which some support exists in earlier literature. For instance, Garcı´a-Carpintero et al. (2010) found a strong positive correlation, for all 15 disciplines taken into account, between number of members in editorial boards for countries and the publication volumes of the countries in the corresponding journals, where only the top 20 journals, according to the Journal Impact Factor, were used in the study. Thus, the nationality composition of journal boards, which is related to the competitiveness force factor, seems to be correlated to the disciplinary structures of countries, at least when it comes to highly cited journals. Two limitations of the study need to be pointed out. First, our results might be affected by changes in database coverage from the first used time period to the second. Second, the four BRICS nations (Brazil, Russia, India, and South Africa) have small numbers of publications in the NSP journals for both periods. Analyses of disciplinary structure based on small sample sizes might suffer from reliability problems. For future research, we would like to study disciplinary structure in Nature, Science, and PNAS based on academic impact, measured by citations, with the use of the same baselines as in the present work. Acknowledgements We thank Ronald Rousseau for valuable comments. We also thank two anonymous reviewers for suggestions that helped us to improve the paper considerably. This work is supported by the National Natural Science Foundation of China (Grant No. L1422060).

Appendix See Tables 6, 7, 8, 9, 10, 11, 12, 13 and 14.

123

Scientometrics Table 6 Number of publications in Nature, Science and PNAS by discipline and time period C15 category

Nature

Science

PNAS

2004–2006

2014–2016

2004–2006

2014–2016

2004–2006

2014–2016

#pub

#pub

#pub

#pub

share (%)

#pub

#pub

share (%)

share (%)

share (%)

share (%)

share (%)

AGRI

1

0.0

3

0.1

4

0.1

0.0

3

0.0

5

0.1

BIOL

1243

44.0

1250

48.1

1071

39.7

912

37.8

5599

59.0

5159

52.6

CHEM

114

4.0

155

6.0

252

9.3

281

11.6

620

6.5

453

4.6

COMP

1

0.0

4

0.2

7

0.3

6

0.2

11

0.1

13

0.1

ECON

0

0.0

6

0.2

5

0.2

19

0.8

12

0.1

63

0.6

ENGI

3

0.1

4

0.2

4

0.1

13

0.5

17

0.2

17

0.2

ENVI

82

2.9

73

2.8

92

3.4

83

3.4

136

1.4

357

3.6

GEOS

366

13.0

221

8.5

336

12.4

197

8.2

122

1.3

411

4.2

MATE

18

0.6

28

1.1

43

1.6

54

2.2

29

0.3

82

0.8

MATH

0

0.0

0

0.0

0

0.0

0

0.0

46

0.5

40

0.4

MEDI

443

15.7

431

16.6

431

16.0

398

16.5

2674

28.2

2568

26.2

PHYS

297

10.5

211

8.1

237

8.8

289

12.0

156

1.6

459

4.7

SOCI

22

0.8

14

0.5

20

0.7

27

1.1

60

0.6

145

1.5

SPAC

233

8.3

198

7.6

199

7.4

135

5.6

12

0.1

43

0.4

Total

2823

100.0

2598

100.0

2701

100.0

2414

100.0

9497

100.0

9815

100.0

Table 7 Number of publications in Non-NSP_Multi and Non-Multi by discipline and time period C15 category

Non-NSP_Multi

Non-Multi

2004–2006

2014–2016

2004–2006

#pub

#pub

#pub

share (%)

share (%)

2014–2016 share (%)

#pub

share (%)

AGRI

96

1.4

1591

1.1

63,750

2.5

121,954

2.9

BIOL

2296

32.8

47,065

33.7

401,486

16.0

589,252

13.8

CHEM

171

2.4

5708

4.1

325,602

13.0

509,958

12.0

COMP

56

0.8

858

0.6

48,924

2.0

114,253

2.7

ECON

9

0.1

554

0.4

36,624

1.5

76,785

1.8

ENGI

166

2.4

1985

1.4

163,207

6.5

383,299

9.0

ENVI

237

3.4

5074

3.6

66,420

2.7

146,352

3.4

GEOS

531

7.6

2792

2.0

75,196

3.0

135,251

3.2

MATE

56

0.8

2330

1.7

117,046

4.7

260,509

6.1

MATH

108

1.5

949

0.7

70,871

2.8

119,615

2.8

MEDI

2813

40.2

58,860

42.2

747,448

29.9

1,226,823

28.8

PHYS

202

2.9

9321

6.7

251,335

10.0

308,235

7.2

SOCI

179

2.6

2339

1.7

102,034

4.1

227,467

5.3

SPAC

79

1.1

137

0.1

32,037

1.3

43,113

1.0

6999

100.0

139,563

100.0

2,501,980

100.0

4,262,866

100.0

Total

123

46.5

57.9

Australia

Netherlands

7.8

11.4

9.2

5.3

51.9

42.7

64.4

70.4

107.9

126.9

31.4

157.5

203.6

1583.8

6.0

20.8

23.1

23.2

113.9

129.1

137.8

218.7

323.4

465.4

88.7

429.6

476.1

6110.2

22.8

46.4

36.7

38.3

223.7

218.3

258.7

365.6

557.0

735.0

145.9

774.6

981.5

9190.0

267.1

56.2

1331.0

187.7

80.9

152.2

68.2

106.8

194.5

384.4

99.1

328.9

334.9

1886.9

Non-NSP_Multi

9005.4

50,687.4

58,873.7

40,407.4

46,371.7

56,724.6

29,433.3

89,298.7

103,369.3

185,811.0

160,993.5

148,174.6

157,643.7

697,282.0

Non-Multi 1

21

18

29

20

7

9

8

6

5

4

11

3

2

22

17

20

25

8

9

7

6

5

4

11

3

2

1

Science

36

24

21

20

12

11

9

6

5

3

13

4

2

1

PNAS

28

19

23

20

9

10

7

6

5

4

13

3

2

1

NSP

7

21

2

9

16

10

19

12

8

3

13

6

4

1

Non-NSP_Multi

35

13

10

15

14

12

19

7

6

2

3

5

4

1

Non-Multi

The 14 countries are selected according to the fractional publication output in NSP in the period 2014–2016 and include the BRICS countries. The countries are sorted descending after NSP total publication output with respect to the period 2014–2016, as in Table 4

8.9

56.5

Switzerland

14.2

76.5

Canada

South Africa

125.7

France

Russia

142.6

Japan

4.4

25.8

China

India

187.5

Germany

9.8

301.8

Brazil

1496.0

UK

NSP

Nature

PNAS

Nature

Science

Rank

Number of publication fractions

USA

Countries

Table 8 Number of publication fractions and ranks of the 14 countries in the time period 2004-2006 by unit of analysis/NSP

Scientometrics

123

123

AUS

–

0.951

0.513

0.778

0.491

0.545

0.851

0.196

0.640

0.410

0.983

0.738

0.709

0.595

2014–2016 2004–2006

AUS

BRA

CAN

CHN

FRA

GER

IND

JAP

NET

RUS

SOU

SWI

UK

USA

0.580

0.651

0.659

0.931

0.536

0.631

0.231

0.808

0.529

0.617

0.711

0.474

–

0.950

BRA

0.944

0.958

0.756

0.412

0.499

0.668

0.873

0.525

0.911

0.621

0.670

–

0.872

0.894

CAN

0.728

0.803

0.962

0.700

0.525

0.702

0.350

0.693

0.726

0.694

–

0.664

0.404

0.379

CHN

0.776

0.654

0.729

0.386

0.920

0.820

0.573

0.580

0.808

–

0.652

0.742

0.561

0.552

FRA

0.953

0.920

0.809

0.430

0.730

0.887

0.825

0.613

–

0.821

0.756

0.859

0.744

0.641

GER

Table 9 Similarity values (cosine measure) between the 14 countries in Nature

0.663

0.695

0.696

0.815

0.517

0.636

0.315

–

0.638

0.916

0.467

0.544

0.465

0.420

IND

0.778

0.760

0.428

0.126

0.577

0.552

–

0.542

0.832

0.741

0.919

0.635

0.349

0.305

JAP

0.808

0.742

0.782

0.547

0.818

–

0.477

0.623

0.790

0.685

0.517

0.915

0.958

0.919

NET

0.638

0.514

0.559

0.356

–

0.520

0.690

0.692

0.871

0.805

0.550

0.573

0.505

0.352

RUS

0.470

0.615

0.633

–

0.419

0.885

0.304

0.776

0.576

0.723

0.329

0.718

0.817

0.822

SOU

0.842

0.846

–

0.551

0.718

0.769

0.878

0.576

0.942

0.815

0.859

0.921

0.696

0.685

SWI

0.939

–

0.954

0.659

0.573

0.821

0.794

0.602

0.848

0.789

0.875

0.929

0.732

0.757

UK

–

0.924

0.912

0.467

0.580

0.637

0.895

0.598

0.808

0.771

0.945

0.779

0.487

0.479

USA

Scientometrics

AUS

–

0.802

0.766

0.703

0.750

0.684

0.619

0.673

0.688

0.767

0.744

0.733

0.904

0.858

2014–2016 2004–2006

AUS

BRA

CAN

CHN

FRA

GER

IND

JAP

NET

RUS

SOU

SWI

UK

USA

0.678

0.809

0.671

0.730

0.590

0.653

0.552

0.601

0.692

0.575

0.464

0.790

–

0.955

BRA

0.735

0.854

0.640

0.778

0.518

0.801

0.585

0.689

0.602

0.552

0.475

–

0.756

0.840

CAN

0.789

0.729

0.870

0.222

0.293

0.781

0.890

0.470

0.808

0.622

–

0.638

0.280

0.344

CHN

0.893

0.774

0.658

0.351

0.693

0.602

0.810

0.545

0.812

–

0.681

0.730

0.373

0.571

FRA

0.886

0.789

0.898

0.226

0.389

0.806

0.946

0.675

–

0.955

0.780

0.801

0.348

0.519

GER

Table 10 Similarity values (cosine measure) between the 14 countries in Science

0.681

0.676

0.722

0.422

0.370

0.547

0.536

–

0.634

0.595

0.587

0.691

0.491

0.575

IND

0.913

0.807

0.892

0.216

0.387

0.877

–

0.700

0.945

0.903

0.832

0.717

0.243

0.395

JAP

0.858

0.883

0.859

0.503

0.424

–

0.916

0.821

0.872

0.821

0.865

0.846

0.549

0.633

NET

0.686

0.726

0.389

0.753

–

0.586

0.774

0.474

0.833

0.851

0.509

0.586

0.152

0.371

RUS

0.507

0.740

0.384

–

0.089

0.299

0.103

0.251

0.311

0.270

0.273

0.453

0.516

0.561

SOU

0.873

0.863

–

0.424

0.604

0.824

0.751

0.519

0.796

0.749

0.910

0.728

0.519

0.579

SWI

0.939

–

0.782

0.681

0.519

0.859

0.748

0.732

0.838

0.764

0.776

0.836

0.548

0.669

UK

–

0.960

0.843

0.545

0.649

0.913

0.876

0.726

0.914

0.863

0.865

0.788

0.426

0.575

USA

Scientometrics

123

123

AUS

–

0.650

0.979

0.450

0.686

0.538

0.541

0.556

0.716

0.416

0.305

0.553

0.829

0.781

2014–2016 2004–2006

AUS

BRA

CAN

CHN

FRA

GER

IND

JAP

NET

RUS

SOU

SWI

UK

USA

0.739

0.748

0.388

0.891

0.219

0.428

0.212

0.774

0.362

0.606

0.725

0.576

–

0.937

BRA

0.779

0.830

0.610

0.205

0.536

0.781

0.563

0.494

0.553

0.706

0.440

–

0.905

0.954

CAN

0.811

0.842

0.665

0.725

0.671

0.549

0.546

0.545

0.609

0.747

–

0.856

0.639

0.689

CHN

0.959

0.918

0.913

0.361

0.793

0.861

0.694

0.771

0.899

–

0.981

0.877

0.706

0.747

FRA

0.867

0.762

0.960

0.182

0.769

0.868

0.881

0.698

–

0.989

0.963

0.902

0.770

0.794

GER

Table 11 Similarity values (cosine measure) between the 14 countries in PNAS

0.800

0.678

0.655

0.614

0.348

0.628

0.411

–

0.825

0.883

0.930

0.678

0.446

0.445

IND

0.738

0.692

0.827

0.098

0.697

0.763

–

0.908

0.892

0.894

0.952

0.807

0.619

0.614

JAP

0.875

0.796

0.944

0.112

0.828

–

0.919

0.822

0.955

0.943

0.953

0.956

0.796

0.839

NET

0.737

0.726

0.890

0.044

–

0.898

0.838

0.836

0.931

0.940

0.922

0.818

0.642

0.667

RUS

0.512

0.529

0.173

–

0.421

0.596

0.289

0.148

0.498

0.451

0.377

0.790

0.842

0.908

SOU

0.889

0.796

–

0.510

0.800

0.942

0.909

0.883

0.914

0.934

0.951

0.899

0.723

0.772

SWI

0.966

–

0.859

0.558

0.931

0.961

0.866

0.749

0.956

0.938

0.932

0.915

0.727

0.815

UK

–

0.978

0.917

0.595

0.861

0.971

0.877

0.755

0.939

0.927

0.935

0.938

0.734

0.848

USA

Scientometrics

AUS

–

0.780

0.511

0.707

0.670

0.574

0.687

0.303

0.681

0.314

0.655

0.542

0.486

0.532

2014–2016 2004–2006

AUS

BRA

CAN

CHN

FRA

GER

IND

JAP

NET

RUS

SOU

SWI

UK

USA

0.533

0.497

0.631

0.550

0.368

0.510

0.516

0.889

0.666

0.737

0.863

0.428

–

0.960

BRA

0.737

0.914

0.470

0.621

0.507

0.516

0.754

0.428

0.656

0.682

0.537

–

0.949

0.976

CAN

0.572

0.449

0.623

0.455

0.699

0.482

0.579

0.725

0.840

0.790

–

0.758

0.754

0.770

CHN

0.898

0.740

0.913

0.622

0.590

0.759

0.676

0.588

0.927

–

0.868

0.860

0.882

0.832

FRA

0.848

0.627

0.857

0.534

0.631

0.718

0.733

0.543

–

0.976

0.930

0.855

0.859

0.824

GER

0.517

0.498

0.489

0.734

0.374

0.422

0.609

–

0.762

0.858

0.704

0.635

0.749

0.659

IND

Table 12 Similarity values (cosine measure) between the 14 countries in Non-NSP_Multi

0.686

0.751

0.596

0.583

0.450

0.349

–

0.743

0.975

0.938

0.945

0.763

0.750

0.716

JAP

0.859

0.633

0.821

0.719

0.273

–

0.758

0.608

0.812

0.825

0.696

0.922

0.818

0.859

NET

0.531

0.374

0.392

0.423

–

0.465

0.742

0.766

0.716

0.763

0.657

0.381

0.436

0.383

RUS

0.777

0.736

0.608

–

0.574

0.808

0.529

0.760

0.631

0.744

0.510

0.774

0.769

0.798

SOU

0.868

0.639

–

0.744

0.753

0.853

0.955

0.835

0.981

0.983

0.923

0.868

0.871

0.852

SWI

0.838

–

0.966

0.863

0.682

0.922

0.869

0.815

0.924

0.959

0.833

0.927

0.895

0.908

UK

–

0.975

0.977

0.797

0.742

0.873

0.912

0.804

0.954

0.965

0.886

0.887

0.877

0.876

USA

Scientometrics

123

123

AUS

–

0.869

0.996

0.565

0.845

0.833

0.697

0.697

0.963

0.609

0.972

0.912

0.963

0.972

2014–2016 2004–2006

AUS

BRA

CAN

CHN

FRA

GER

IND

JAP

NET

RUS

SOU

SWI

UK

USA

0.886

0.855

0.976

0.801

0.765

0.887

0.920

0.881

0.961

0.947

0.780

0.889

–

0.866

BRA

0.960

0.949

0.927

0.967

0.623

0.954

0.734

0.737

0.847

0.859

0.613

–

0.921

0.987

CAN

0.570

0.566

0.763

0.496

0.768

0.569

0.917

0.968

0.796

0.792

–

0.712

0.782

0.636

CHN

0.859

0.844

0.978

0.782

0.902

0.857

0.924

0.899

0.989

–

0.847

0.847

0.858

0.781

FRA

0.872

0.855

0.976

0.742

0.870

0.879

0.945

0.896

–

0.989

0.855

0.885

0.896

0.828

GER

Table 13 Similarity values (cosine measure) between the 14 countries in Non-Multi

0.704

0.693

0.878

0.641

0.851

0.704

0.946

–

0.895

0.886

0.979

0.713

0.804

0.636

IND

0.731

0.715

0.901

0.589

0.830

0.739

–

0.939

0.957

0.942

0.908

0.760

0.852

0.674

JAP

0.993

0.985

0.918

0.888

0.627

–

0.717

0.618

0.854

0.797

0.590

0.958

0.851

0.967

NET

0.647

0.648

0.813

0.575

–

0.542

0.875

0.849

0.868

0.902

0.796

0.597

0.622

0.525

RUS

0.909

0.898

0.853

–

0.566

0.926

0.648

0.645

0.819

0.787

0.618

0.962

0.853

0.977

SOU

0.912

0.889

–

0.875

0.808

0.890

0.911

0.847

0.982

0.970

0.826

0.943

0.920

0.891

SWI

0.996

–

0.908

0.947

0.651

0.978

0.744

0.669

0.889

0.846

0.630

0.947

0.835

0.959

UK

–

0.989

0.944

0.940

0.658

0.984

0.799

0.713

0.922

0.881

0.677

0.969

0.897

0.961

USA

Scientometrics

- 2.1

- 0.6

- 12.0

- 0.1

1.8

- 3.1

- 3.2

0.4

- 0.2

- 0.3

2.0

5.0

25.8

BRA

CAN

CHN

FRA

GER

IND

JAP

NET

RUS

SOU

SWI

UK

USA

- 3.2

- 3.0

IND

JAP

FRA

1.7

0.6

CHN

GER

- 1.1

- 13.6

CAN

- 1.7

- 2.3

AUS

BRA

Science

- 1.1

AUS

Nature

- 0.7

- 3.6

1.9

0.4

- 26.2

1.5

- 1.3

- 1.8

39.0

3.7

2.2

- 0.3

- 0.3

0.2

- 0.5

- 3.7

4.6

- 0.5

- 25.0

- 0.1

- 1.2

- 1.9

5.5

- 1.7

- 3.1

- 1.1

- 0.6

- 16.0

2.1

- 0.9

2.3

4.4

7.5

2.3

- 0.5

0.1

4.6

- 2.0

- 3.1

6.0

0.8

- 16.0

1.2

- 2.6

- 0.2

- 10.7

2.2

0.7

- 15.6

- 0.1

- 3.0

- 1.6

27.3

5.4

2.8

- 1.1

- 0.3

0.0

- 1.4

- 10.9

1.6

2.5

- 15.1

0.9

- 2.8

1.7

GEOS

- 0.7

- 2.2

0.9

1.4

- 23.2

- 0.9

- 0.6

- 2.3

36.0

5.5

0.8

- 0.1

- 0.5

- 0.8

- 1.5

- 2.2

- 1.5

- 0.4

- 23.8

- 1.1

- 0.8

- 2.6

MATE

- 1.3

- 1.2

0.5

- 0.1

- 13.7

- 1.1

- 1.4

- 1.4

42.2

2.3

1.2

- 0.4

- 0.1

- 1.7

- 4.7

- 1.5

- 0.7

0.0

- 15.1

- 0.7

- 2.0

- 1.5

MEDI

- 1.4

- 1.8

4.9

1.3

- 22.0

2.3

- 0.8

- 1.2

18.6

1.1

2.6

- 0.2

0.0

0.4

- 2.5

- 1.7

9.8

2.8

- 22.2

0.6

- 0.4

- 0.6

PHYS

- 1.9

- 3.3

0.5

- 1.3

- 6.1

- 4.0

- 2.3

- 5.0

–

–

–

–

–

–

–

–

–

–

–

–

–

–

SOCI

0.1

- 9.4

5.0

4.5

- 14.2

1.5

- 0.3

2.7

17.5

1.2

0.5

- 1.1

- 1.8

1.2

- 2.1

- 9.0

2.7

6.7

- 12.0

1.0

- 0.7

1.0

SPAC

- 1.7

- 3.4

2.5

1.3

- 10.9

- 1.0

- 3.7

- 1.3

27.6

6.1

2.2

- 0.5

- 0.9

0.5

- 1.9

- 3.4

2.6

0.6

- 9.3

- 0.6

- 3.5

- 0.7

0.1

- 6.5

1.3

- 0.6

- 21.8

1.5

- 1.6

- 0.9

42.0

4.7

2.6

- 0.4

- 3.3

0.6

0.2

- 6.6

3.9

- 1.4

- 20.6

- 0.1

- 1.5

- 0.9

CHEM

BIOL

ENVI

BIOL

CHEM

Non-Multi

Non-NSP_Multi

- 1.6

- 2.9

- 0.5

- 0.8

- 13.6

1.8

0.5

4.6

7.1

8.7

2.4

0.1

- 0.6

4.5

- 1.9

- 3.0

6.6

0.5

- 13.6

0.9

- 1.2

7.8

ENVI

0.9

- 2.7

1.5

0.0

- 14.6

- 1.1

- 1.3

- 1.3

23.7

6.3

2.9

- 0.5

- 3.6

- 0.1

- 0.3

- 2.9

1.0

1.9

- 14.1

- 0.1

- 1.1

2.0

GEOS

- 0.6

- 5.7

0.4

0.3

- 23.7

- 1.2

- 1.1

- 1.4

44.4

6.5

1.2

- 0.3

- 2.0

- 0.6

- 1.5

- 5.7

- 1.9

- 1.6

- 24.4

- 1.3

- 1.3

- 1.6

MATE

0.0

- 1.5

0.9

- 0.3

- 5.6

- 1.5

- 1.4

- 1.8

35.2

2.0

1.4

- 0.3

- 0.3

- 1.0

- 3.4

- 1.8

- 0.2

- 0.2

- 7.0

- 1.0

- 2.0

- 1.9

MEDI

- 1.3

- 4.6

5.2

0.5

- 16.7

2.2

- 1.4

- 0.5

22.3

2.7

2.9

- 0.3

- 4.5

0.6

- 2.3

- 4.5

10.1

2.0

- 16.9

0.5

- 1.0

0.1

PHYS

- 0.9

- 0.7

- 0.3

- 0.3

- 1.2

- 4.4

- 1.8

- 5.5

–

–

–

–

–

–

–

–

–

–

–

–

–

–

SOCI

0.8

- 3.2

2.7

3.0

- 6.7

- 0.4

- 0.7

1.0

17.0

- 0.2

0.9

0.3

- 3.9

0.6

- 1.3

- 2.8

0.4

5.3

- 4.5

- 0.9

- 1.1

- 0.7

SPAC

Table 14 Differences within countries between NSP journals and the two baselines with respect to world share of publications. Only main disciplines considered (cf. last paragraph of Sect. 3.2). Publication period = 2014–2016

Scientometrics

123

123

- 0.1

- 0.3

- 0.3

1.0

4.6

30.7

NET

RUS

SOU

SWI

UK

USA

MATE

MEDI

PHYS

SOCI

SPAC

- 0.5

- 12.3

- 0.1

- 0.3

- 2.8

- 1.6

- 0.1

- 0.3

- 0.4

0.4

1.2

31.5

CAN

CHN

FRA

GER

IND

JAP

NET

RUS

SOU

SWI

UK

USA

1.3

44.7

0.5

1.5

- 0.3

- 0.3

0.3

- 2.5

- 3.0

0.3

1.9

- 25.1

0.4

- 1.3

- 1.3

42.4

4.0

- 0.4

- 0.3

- 0.4

0.6

32.3

0.1

0.6

- 0.5

0.1

0.4

- 0.8

- 3.1

- 0.2

0.0

- 15.3

1.1

- 3.2

- 1.6

17.4

4.5

2.6

- 1.0

- 0.1

37.4

1.2

0.1

- 1.1

- 0.2

0.2

- 2.9

- 10.7

3.6

1.7

- 14.0

0.0

- 2.6

- 1.0

39.2

0.8

0.9

- 1.3

0.5

0.4

1.1

–

–

–

–

–

–

–

–

–

–

–

–

–

–

30.7

0.7

5.6

- 0.1

- 0.3

36.0

1.1

0.2

- 0.4

0.0

- 1.3

- 1.9

- 1.2

- 0.2

0.0

- 14.0

- 0.5

- 1.7

- 1.0

34.1

2.9

1.2

- 0.5

- 0.1

- 0.3

0.5

28.9

2.3

0.3

- 0.2

- 0.9

1.0

- 2.3

- 0.9

0.6

2.0

- 19.6

0.9

- 0.7

- 1.3

24.2

0.5

2.7

- 0.2

- 0.4

40.8

- 0.6

- 0.8

- 2.8

0.0

- 2.5

- 0.9

- 3.3

0.5

- 0.4

- 5.4

- 2.4

- 2.3

- 3.8

35.1

1.6

0.8

- 1.9

- 0.2

- 3.5

–

–

–

–

–

–

–

–

–

–

–

–

–

–

0.5

21.5

- 1.5

0.3

- 2.1

- 1.7

0.0

33.3

2.3

0.5

- 0.6

- 1.0

0.0

- 0.3

- 3.1

0.5

0.6

- 9.6

- 0.5

- 3.4

- 1.1

32.5

5.7

1.1

- 0.5

- 1.0

47.7

1.5

1.9

- 0.4

- 3.3

0.7

- 1.8

- 5.9

- 0.3

1.0

- 20.6

0.3

- 1.6

- 0.4

45.4

5.0

0.0

- 0.4

- 3.4

1.7

CHEM

34.9

1.3

0.7

0.1

- 0.7

0.3

- 0.6

- 3.0

0.4

- 0.3

- 12.9

0.8

- 1.8

0.7

20.1

5.7

2.7

- 0.3

- 0.8

0.4

ENVI

33.8

2.0

0.1

- 0.4

- 3.5

0.1

- 1.8

- 2.7

3.0

1.0

- 13.1

- 0.9

- 0.8

- 0.7

35.5

1.7

1.0

- 0.6

- 2.7

0.3

GEOS

–

–

–

–

–

–

–

–

–

–

–

–

–

–

39.1

1.7

5.9

- 0.3

- 1.8

1.3

MATE

29.1

0.8

0.4

- 0.3

- 0.3

- 0.6

- 0.5

- 1.6

0.3

- 0.2

- 5.9

- 0.9

- 1.7

- 1.5

27.1

2.6

1.4

- 0.3

- 0.3

0.4

MEDI

32.6

3.9

0.7

- 0.3

- 5.3

1.1

- 2.1

- 3.7

0.9

1.2

- 14.3

0.8

- 1.3

- 0.6

28.0

2.0

3.0

- 0.3

- 4.9

0.7

PHYS

29.1

- 2.1

- 0.5

- 1.0

0.0

- 1.9

0.1

- 0.7

- 0.3

0.6

- 0.5

- 2.9

- 1.8

- 4.4

23.4

0.2

1.0

- 0.2

- 0.3

- 2.8

SOCI

–

–

–

–

–

–

–

–

–

–

–

–

–

–

21.0

- 2.8

0.7

- 0.7

- 3.7

- 0.1

SPAC

The values in the table cells are country world shares of publications in Nature/Science/PNAS minus corresponding shares in Non-NSP_Multi/Non-Multi. The countries marked in italics are the five countries explicitly mentioned in the last paragraph of Sect. 3.2.1

- 1.4

- 2.1

AUS

BRA

PNAS

GEOS

BIOL

ENVI

BIOL

CHEM

Non-Multi

Non-NSP_Multi

Table 14 continued

Scientometrics

Scientometrics

References Aksnes, D. W., Leeuwen, T. N. V., & Sivertsen, G. (2014). The effect of booming countries on changes in the relative specialization index (rsi) on country level. Scientometrics, 101, 1391–1401. Bongioanni, I., Daraio, C., & Ruocco, G. (2014). A quantitative measure to compare the disciplinary profiles of research systems and their evolution over time. Journal of Informetrics, 8(3), 710–727. Bo¨rner, K., Klavans, R., Patek, M., Zoss, A. M., Biberstine, J. R., Light, R. P., et al. (2012). Design and update of a classification system: The UCSD map of science. PLoS ONE, 7(7), e39464. Boyack, K. W., Klavans, R., & Bo¨rner, K. (2005). Mapping the backbone of science. Scientometrics, 64(3), 351–374. Braun, T., Gla¨nzel, W., & Schubert, A. (1989). National publication patterns and citation impact in the multidisciplinary journals Nature and Science. Scientometrics, 17(1–2), 11–14. Ding, J. L., Ahlgren, P., Yang, L. Y., & Yue, T. (2016). Document type profiles in Nature, Science and PNAS: Journal and country level. Journal of Data and Information Science, 1(3), 27–41. Ding, J. L., & Rousseau, R. (2015). Do Chinese and American contributions in top journals have an equal citation potential? Chinese Journal of Library and Information Science, 8(2), 1–10. Garcı´a-Carpintero, E., Granadino, B., & Plaza, L. M. (2010). The representation of nationalities on the editorial boards of international journals and the promotion of the scientific output of the same countries. Scientometrics, 84(3), 799–811. Gla¨nzel, W. (2000). Science in Scandinavia: A bibliometric approach. Scientometrics, 48(2), 121–150. Gla¨nzel, W., Debackere, K., & Meyer, M. (2008). ‘Triad’ or ‘tetrad’? On global changes in a dynamic world. Scientometrics, 74(1), 71–88. Gla¨nzel, W., & Schubert, A. (2003). A new classification scheme of science fields and subfields designed for scientometric evaluation purposes. Scientometrics, 56(3), 357–367. Gla¨nzel, W., Schubert, A., & Czerwon, J. (1999a). An Item-by-item subject classification of papers published in multidisciplinary and general journals using reference analysis. Scientometrics, 44(3), 427–439. Gla¨nzel, W., Schubert, A., Schoepflin, U., & Czerwon, J. (1999b). An item-by-item subject classification of papers published in journals covered by the SSCI database using reference analysis. Scientometrics, 46(3), 431–441. Harzing, A. W., & Giroud, A. (2014). The competitive advantage of nations: An application to academia. Journal of Informetrics, 8, 29–42. Jurajda, S., Kozubek, S., Mu¨nich, D., & Sˇkoda, S. (2017). Scientific publication performance in postcommunist countries: Still lagging far behind. Scientometrics, 112(1), 315–328. Kaneiwa, K., Adachi, J., Aoki, M., Masuda, T., Midorikawa, N., Tanimura, A., et al. (1988). A comparison between the journals Nature and Science. Scientometrics, 13(3–4), 125–153. King, D. A. (2004). The scientific impact of nations. Nature, 430(6997), 311–316. Klavans, R., & Boyack, K. W. (2006). Identifying a better measure of relatedness for mapping science. Journal of the American Society for Information Science and Technology, 57(2), 251–263. Klavans, R., & Boyack, K. W. (2009). Toward a consensus map of science. Journal of the American Society for Information Science and Technology, 60(3), 455–476. Kozlowski, J., Radosevic, S., & Ircha, D. (1999). History matters: The inherited disciplinary structure of the post-communist science in countries of central and Eastern Europe and its restructuring. Scientometrics, 45(1), 137–166. Levenshtein, V. I. (1966). Binary codes capable of correcting deletions, insertions and reversals. Soviet Physics Doklady, 10(8), 707–710. Leydesdorff, L., Carley, S., & Rafols, I. (2013). Global maps of science based on the new Web-of-Science categories. Scientometrics, 94(2), 589–593. Leydesdorff, L., & Rafols, I. (2009). A global map of science based on the ISI subject categories. Journal of the American Society for Information Science and Technology, 60(2), 348–362. Li, F., Miao, Y., & Ding, J. (2015). Tracking the development of disciplinary structure in China’s top research universities (1998–2013). Research Evaluation, 24(3), 312–324. Li, N. (2017). Evolutionary patterns of national disciplinary profiles in research: 1996–2015. Scientometrics, 111, 493–520. Moya-Anego´n, F., & Herrero-Solana, V. (2013). Worldwide topology of the scientific subject profile: A macro approach in the country level. PLoS ONE, 8(12), e83222. ´ lvarez, E., Munoz-Ferna´ndez, Moya-Anego´n, S. G., Vargas-Quesada, B., Chinchilla-Rodrı´guez, Z., Corera-A F. J., & Herrero-Solana, V. (2007). Visualizing the marrow of science. Journal of the American Society for Information Science and Technology, 58(14), 2167–2179.

123

Scientometrics ´ lvarez, E., Moya-Anego´n, F., Vargas-Quesada, B., Herrero-Solana, V., Chinchilla-Rodrı´guez, Z., Corera-A & Munoz-Ferna´ndez, F. J. (2004). A new technique for building maps of large scientific domains based on the cocitation of classes and categories. Scientometrics, 61(1), 129–145. OECD. (2018). Main science and technology indicators volume 2017 Issue 2, OECD Publishing, Paris. http://dx.doi.org/10.1787/msti-v2017-2-en. Porter, A., & Rafols, I. (2009). Is science becoming more interdisciplinary? Measuring and mapping six research fields over time. Scientometrics, 81(3), 719–745. Porter, A. L., Roessner, D. J., & Heberger, A. E. (2008). How interdisciplinary is a given body of research? Research Evaluation, 17(4), 273–282. Radosevic, S., & Yoruk, E. (2014). Are there global shifts in the world science base? Analysing the catching up and falling behind of world regions. Scientometrics, 101(3), 1897–1924. Rafols, I., Porter, A. L., & Leydesdorff, L. (2010). Science overlay maps: A new tool for research policy and library management. Journal of the American Society for Information Science and Technology, 61(9), 1871–1887. Rousseau, R., & Ding, J. L. (2016). Does international collaboration yield a higher citation potential for US scientists publishing in highly visible interdisciplinary Journals? Journal of the Association for Information Science and Technology, 57(4), 1009–1013. Salton, G., & McGill, M. J. (1983). Introduction to modern information retrieval. New York: McGraw-Hill Inc. Schwab, K. (2017). The global competitiveness report 2016–2017. Geneva: World Economic Forum. http:// www3.weforum.org/docs/GCR2017-2018/05FullReport/TheGlobalCompetitivenessReport2017%E2% 80%932018.pdf. Van den Besselaar, P., & Heimeriks, G. (2001). Disciplinary, multidisciplinary, interdisciplinary: Concepts and indicators. In M. Davis & C. S. Wilson (Eds.), Proceedings of the 8th International Conference on Scientometrics and Informetrics (pp. 705–716). Sydney: University of New South-Wales. Van Eck, N. J., & Waltman, L. (2010). Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics, 84(2), 523–538. Vinkler, P. (2018). Structure of the scientific research and science policy. Scientometrics, 114(2), 737–756. Waltman, L., & Van Eck, N. J. (2012). A new methodology for constructing a publication-level classification system of science. Journal of the American Society for Information Science and Technology, 63(12), 2378–2392. Waltman, L., Van Eck, N. J., & Noyons, E. (2010). A unified approach to mapping and clustering of bibliometric networks. Journal of Informetrics, 4(4), 629–635. Wang, Q. (2016). Studies in the dynamics of science: Exploring emergence, classification, and interdisciplinarity. Stockholm: KTH Royal institute of Technology. Wang, X. W., Mao, W. L., Xu, S. M., & Zhang, C. B. (2014). Usage history of scientific literature: Nature metrics and metrics of Nature publications. Scientometrics, 98, 1923–1933. Wong, C. Y. (2013). On a path to creative destruction: science, technology and science-based technological trajectories of Japan and South Korea. Scientometrics, 96(1), 323–336. Xie, Z., Li, M., Li, J., Duan, X., & Ouyang, Z. (2018). Feature analysis of multidisciplinary scientific collaboration patterns based on pnas. EPJ Data Science, 7, 5. Yang, L. Y., Yue, T., Ding, J. L., & Han, T. (2012). A comparison of disciplinary structure in science between the G7 and the BRIC countries by bibliometric methods. Scientometrics, 93, 497–516. Zhou, P., & Glanzel, W. (2010). In-depth analysis on China’s international cooperation in science. Scientometrics, 82(3), 597–612.

123